Selected work

Representative Papers

STINER: Automated Extraction of Strategic Cyber Threat Intelligence from X

Aug 14, 2026

This study addresses the challenge of extracting strategic-level Cyber Threat Intelligence (CTI) from informal social media text. We propose a fine-grained taxonomy tailored for strategic intelligence and construct the first expert-annotated threat alert corpus. Furthermore, we introduce a novel entity recognition framework integrating a DarkBERT domain-adaptive encoder with generative large language models. Experimental results demonstrate that this approach achieves an F1-score of 89.33%, significantly outperforming both general-purpose models and standalone LLMs. Notably, the system successfully provided early warnings for the SafePay ransomware campaign, validating its effectiveness and practical utility in efficiently extracting strategic CTI from complex, unstructured social media content.

0 citationsRead paper

Graph Classification via Network Usable Information: From Representation Evaluation to Structure Selection

Jul 03, 2026

This work addresses the limited interpretability and difficulty in evaluating representation quality inherent in traditional graph neural networks that rely on black-box embeddings. The authors propose NetinfoGC, a novel framework that introduces the Network Utility Information (NUI) paradigm to graph classification for the first time. By integrating classical structural descriptors through a propagation mechanism, NetinfoGC constructs permutation-invariant, training-free graph representations. The effectiveness of these representations is evaluated via clustering consistency, and sparse group LASSO is employed to automatically select salient features. Extensive experiments demonstrate that NUI-based classical centrality measures achieve performance on par with or superior to learned representations across multiple benchmarks. Moreover, NUI estimates exhibit strong correlation with downstream classification accuracy, substantially enhancing both model interpretability and computational efficiency.

0 citationsRead paper

Towards an Agent-First Web: Redesigning the Web for AI Agents

Jun 17, 2026

This work addresses the limitations of traditional human-centric web architectures, which hinder legitimate, efficient, and trustworthy interactions for AI agents. The paper proposes the first holistic Agent-First Web framework that treats AI agents as first-class citizens rather than mere crawlers, grounded in ten core design principles. It introduces HTTP metadata for agent identity verification and permission inheritance, an intent-based layered economic model featuring token subscriptions and delegated content mechanisms, and a novel Agent Text Markup Language (ATML) coupled with cryptographic provenance chains to ensure content authenticity. This framework effectively mitigates agent blocking and curbs recursive knowledge distortion, thereby establishing both technical and economic foundations for a collaborative human-agent information ecosystem.

0 citationsRead paper

Toward a Modular Architecture for Embedded AI Agent Systems at the Edge

Jun 01, 2026

Deploying AI agents with sophisticated reasoning and tool-use capabilities on resource-constrained embedded microcontrollers presents significant challenges in memory, energy consumption, and offline operation. This work proposes a modular embedded agent reference architecture that decouples on-device and cloud-augmented intelligence through a hierarchical design. It integrates deterministic real-time control with lightweight inference mechanisms—including compressed neural networks, rule-based logic, and small language models—and incorporates a governance layer to enable observability, policy enforcement, and security management across distributed device fleets. The architecture systematically balances latency, energy efficiency, and reliability, offering a deployable paradigm for edge AI agents that ensures low latency, strong privacy preservation, and scalability.

0 citationsRead paper

Quantum encodings that preserve persistent homology

May 27, 2026

This work addresses a central challenge in quantum topological data analysis (TDA): how to effectively preserve topological invariants—particularly persistent homology—when encoding classical datasets into quantum states. The authors propose a direct quantum encoding approach that bypasses the conventional construction of simplicial complexes, instead focusing on admissible encoding strategies derived directly from raw distance data. By integrating tools from algebraic topology, metric geometry, and quantum information theory, they systematically analyze how different encoding schemes affect persistent homology structures and identify several quantum encodings capable of efficiently preserving essential topological features. This framework substantially reduces computational resource requirements, offering a new paradigm for low-complexity quantum TDA.

0 citationsRead paper

Recent publications

Latest Papers

STINER: Automated Extraction of Strategic Cyber Threat Intelligence from X

Aug 14, 2026

This study addresses the challenge of extracting strategic-level Cyber Threat Intelligence (CTI) from informal social media text. We propose a fine-grained taxonomy tailored for strategic intelligence and construct the first expert-annotated threat alert corpus. Furthermore, we introduce a novel entity recognition framework integrating a DarkBERT domain-adaptive encoder with generative large language models. Experimental results demonstrate that this approach achieves an F1-score of 89.33%, significantly outperforming both general-purpose models and standalone LLMs. Notably, the system successfully provided early warnings for the SafePay ransomware campaign, validating its effectiveness and practical utility in efficiently extracting strategic CTI from complex, unstructured social media content.

0 citationsRead paper

Graph Classification via Network Usable Information: From Representation Evaluation to Structure Selection

Jul 03, 2026

This work addresses the limited interpretability and difficulty in evaluating representation quality inherent in traditional graph neural networks that rely on black-box embeddings. The authors propose NetinfoGC, a novel framework that introduces the Network Utility Information (NUI) paradigm to graph classification for the first time. By integrating classical structural descriptors through a propagation mechanism, NetinfoGC constructs permutation-invariant, training-free graph representations. The effectiveness of these representations is evaluated via clustering consistency, and sparse group LASSO is employed to automatically select salient features. Extensive experiments demonstrate that NUI-based classical centrality measures achieve performance on par with or superior to learned representations across multiple benchmarks. Moreover, NUI estimates exhibit strong correlation with downstream classification accuracy, substantially enhancing both model interpretability and computational efficiency.

0 citationsRead paper

Towards an Agent-First Web: Redesigning the Web for AI Agents

Jun 17, 2026

This work addresses the limitations of traditional human-centric web architectures, which hinder legitimate, efficient, and trustworthy interactions for AI agents. The paper proposes the first holistic Agent-First Web framework that treats AI agents as first-class citizens rather than mere crawlers, grounded in ten core design principles. It introduces HTTP metadata for agent identity verification and permission inheritance, an intent-based layered economic model featuring token subscriptions and delegated content mechanisms, and a novel Agent Text Markup Language (ATML) coupled with cryptographic provenance chains to ensure content authenticity. This framework effectively mitigates agent blocking and curbs recursive knowledge distortion, thereby establishing both technical and economic foundations for a collaborative human-agent information ecosystem.

0 citationsRead paper

Toward a Modular Architecture for Embedded AI Agent Systems at the Edge

Jun 01, 2026

Deploying AI agents with sophisticated reasoning and tool-use capabilities on resource-constrained embedded microcontrollers presents significant challenges in memory, energy consumption, and offline operation. This work proposes a modular embedded agent reference architecture that decouples on-device and cloud-augmented intelligence through a hierarchical design. It integrates deterministic real-time control with lightweight inference mechanisms—including compressed neural networks, rule-based logic, and small language models—and incorporates a governance layer to enable observability, policy enforcement, and security management across distributed device fleets. The architecture systematically balances latency, energy efficiency, and reliability, offering a deployable paradigm for edge AI agents that ensures low latency, strong privacy preservation, and scalability.

0 citationsRead paper

Quantum encodings that preserve persistent homology

May 27, 2026

This work addresses a central challenge in quantum topological data analysis (TDA): how to effectively preserve topological invariants—particularly persistent homology—when encoding classical datasets into quantum states. The authors propose a direct quantum encoding approach that bypasses the conventional construction of simplicial complexes, instead focusing on admissible encoding strategies derived directly from raw distance data. By integrating tools from algebraic topology, metric geometry, and quantum information theory, they systematically analyze how different encoding schemes affect persistent homology structures and identify several quantum encodings capable of efficiently preserving essential topological features. This framework substantially reduces computational resource requirements, offering a new paradigm for low-complexity quantum TDA.

0 citationsRead paper